When i train an AutoML model in Vertex AI, the type of the feature is "numeric", but when i deploy the model, the type is showing as "text" and the model expects a text value to be passed for prediction.
In this case, i created a dataset with {time = X} and {customer satisfaction = y} as the model inputs. I used the transformation to change the type of "time" from "automatic" to "numeric". After deploying the model, when i try to pass "time" as a numeric, i was getting an error. I then passed it as a string and the model worked for me. I used Python in Cloud Shell for this.
{"time" : 10} won't work and gives an error: "Column: time. Error: Expected string_value but got number_value. Type casting is not allowed."
{"time": "10"} works
I'm unsure if X has always to be a string in AutoML classification models, because my understanding is that logistic regression can take numeric values as the X to build a multiclass classification model.
